Meng (Ethan) Li comments

Results 31 comments of


                                            Meng (Ethan) Li

Add --megascale_abort_on_hangs flag for multi-slice TPU jobs

Replaced by https://github.com/apple/axlearn/pull/1010 according to latest recommendations

Add --megascale_abort_on_hangs flag for multi-slice TPU jobs

I closed this PR since https://github.com/apple/axlearn/pull/1010 has been merged

Pathways: pass XLA flags correctly

> @Ethanlm Thanks for reviewing and approving! Do we need anyone else to approve this? Yes code owner approvals are required

Pathways: pass XLA flags correctly

I will test it in our internal env as a final validation

Pathways: pass XLA flags correctly

Let me try again

Fix submission of Dataflow jobs

Closing this PR due to inactivity. Please re-open or file a new PR if this is still important.

Fix submission of Dataflow jobs

Let me ping @zhiyun to take a look. Thank you!

Alternative method of submitting jobs to DF Runner

Closing this PR due to inactivity. Please re-open or file a new PR if this is still important.

Use lambda instead of functools.partial to create single-host config maker

I have not encountered this issue. How can I reproduce it?

Use lambda instead of functools.partial to create single-host config maker

> > I have not encountered this issue. How can I reproduce it? > > This is only triggered if using one of the *-single-host configs, e.g. `fuji-3B-v3-flash-single-host` Yeah I...